pab1it0/prometheus-mcp-server

Docker: docker run -i --rm -e PROMETHEUS_URL="http://your-prometheus:9090" ghcr.io/pab1it0/prometheus-mcp-server:latest
Claude Code: claude mcp add prometheus --env PROMETHEUS_URL=http://your-prometheus:9090 -- docker run -i --rm -e PROMETHEUS_URL ghcr.io/pab1it0/prometheus-mcp-server:latest
Helm: helm install prometheus-mcp-server oci://ghcr.io/pab1it0/charts/prometheus-mcp-server --version 1.1.1 --set prometheus.url="http://prometheus:9090"
Also installable via Docker Desktop MCP Catalog/MCP Toolkit

Usage

• •Set PROMETHEUS_URL (required); optional auth via PROMETHEUS_USERNAME/PASSWORD, PROMETHEUS_TOKEN, or client certs (PROMETHEUS_CLIENT_CERT/KEY)
• •Configure transport via PROMETHEUS_MCP_SERVER_TRANSPORT (stdio default, or http/sse; bind host/port and stateless HTTP for multi-replica)
• •Use TOOL_PREFIX to run multiple instances against different environments (e.g. staging_execute_query)
• •Tools are configurable - you can disable tools you don't use to save context window

Key features

• ✓Tools: health_check, execute_query (PromQL instant), execute_range_query (range with start/end/step), list_metrics (pagination/filtering), get_metric_metadata, get_targets (server-side state/scrape_pool filtering)
• ✓Authentication: basic auth, bearer token, and mutual TLS
• ✓Docker containerization plus an official Helm chart for Kubernetes
• ✓Configurable tool list; PROMETHEUS_DISABLE_LINKS to save context tokens; request timeout default 30s (DDoS protection)

Best for

Letting AI coding assistants query and analyze Prometheus metrics directly from the chat/editor.

Caveats

• ⚠Requires a Prometheus server reachable from the environment
• ⚠Docker/Helm-first distribution; Python 3.10+ if running from source
• ⚠HTTP/SSE transport is opt-in via PROMETHEUS_MCP_SERVER_TRANSPORT; default is stdio